In a landmark moment for global AI governance, the United States, European Union and United Kingdom have signed the Council of Europe’s framework convention on artificial intelligence and human rights, democracy, and the rule of law (the “Convention”), the first legally binding international treaty on AI.
The agreement marks a critical step towards a level of harmonization between global regulatory approaches to AI technologies. The rapid development of AI systems has posed a challenge for policymakers, who are balancing the need to ensure sufficient guardrails are in place for ethical and safe development and deployment of AI, with creating an environment for innovation.
By committing to shared standards, the U.S., EU and UK have set a baseline for the minimum expectations in relation to fundamental areas such as human rights, the democratic process and the rule of law. In particular, parties to the Convention agree to adopt or maintain measures that seek to:
- Protect democratic processes;
- Respect human dignity and individual autonomy;
- Ensure adequate transparency and oversight requirements, including with regard to identifying AI-produced content;
- Ensure accountability and responsibility for adverse impacts on human rights, democracy and the rule of law;
- Ensure equality is respected and discrimination is prohibited;
- Aim at overcoming inequalities;
- Ensure privacy rights and personal data are protected; and
- Promote the reliability of AI systems and trust in their outputs.
Signatories also agree to take steps to establish, as appropriate, controlled environments for developing and testing AI and encouraging and promoting adequate digital literacy and digital skills for all segments of the population. The Convention includes a reporting mechanism, which requires periodic reporting of the steps taken by a jurisdiction to give effect to the Convention’s requirements.
However, the Convention does not grant any enforcement powers, or include penalties for failure to adhere to its requirements. The European Data Protection Supervisor (among others) has called this a “missed opportunity” to lay down a strong and effective AI legal framework with clear and strong safeguards for persons affected by AI systems, calling the Convention “largely declarative [in] nature” with a very high level of generality that could lead to divergent application.
The Convention follows on from the Bletchley Declaration, following the AI Safety Summit hosted in the UK in late 2023, which established a shared framework for mitigating the existential risks posed by advanced “frontier” AI models, emphasizing the need for greater international collaboration to ensure AI systems are safe, secure, and aligned with human values. (See our earlier briefing here.) Ministers of the G7 countries also recently affirmed the global nature of the AI regulatory debate, acknowledging the need to ensure that the benefits of digital technologies are shaped and shared equitably.
While agreement has been reached on core fundamental matters via the Convention, there remains significant global divergence in AI regulatory policy. In the EU, for example, we are now in the transition period before the groundbreaking AI Act’s operative provisions come into effect. (See our breakdown of the AI Act here.) Meanwhile, California has proposed an AI bill that would impose regulations on AI models that reach certain thresholds, including a requirement for developers to ensure the capability to effect a “full shutdown” if required. If signed into law, California’s bill could influence federal AI legislative policy in the U.S., which is high on the agenda following President Biden’s 2023 Executive Order on AI. Meanwhile, the UK is yet to deviate materially from the “pro-innovation” approach implemented by the previous Conservative Government (notwithstanding expectations that a comprehensive AI Bill would be announced by now by the new Labour Government).
While the latest step towards global alignment via the Convention is therefore a further welcomed development in terms of creating more certainty and consistency, it does not eliminate the issues currently faced by global companies grappling with a patchwork of evolving AI laws.
RELATED ARTICLES
EU Reaches Agreement on New “AI Act”: The World’s First Comprehensive AI Law
Key Takeaways from the UK’s AI Summit: The Bletchley Declaration